/
Navigation
Chronicles
Browse all articles
Explore
Semantic exploration
Research
Entity momentum
Nexus
Correlations & relationships
Story Arc
Topic evolution
Drift Map
Semantic trajectory animation
Posts
Analysis & commentary
Pulse API
Tech news intelligence API
Browse
Entities
Companies, people, products, technologies
Domains
Browse by publication source
Handles
Browse by social media handle
Detection
Concept Search
Semantic similarity search
High Impact Stories
Top coverage by position
Sentiment Analysis
Positive/negative coverage
Anomaly Detection
Unusual coverage patterns
Analysis
Rivalry Report
Compare two entities head-to-head
Semantic Pivots
Narrative discontinuities
Crisis Response
Event recovery patterns
Connected
Search: /
Command: ⌘K
Embeddings: large
TEXXR

Chronicles

The story behind the story

days · browse · Enter similar · o open

Source: Microsoft's GitHub Copilot, the $10/month AI assistant with 1.5M+ users, loses an average of $20+/month per user and as much as $80/month for some users

Wall Street Journal

Context & Ripple Effects

GitHub moved Copilot out of preview as a $10-per-month coding assistant after more than 1.2 million people had tried it, making low-friction subscription pricing central to its early commercial model. This report puts a concrete cost constraint behind that model.

Later coverage shows Microsoft seeking other ways to close the gap: bundling Copilot into Microsoft 365 alongside a price increase and, eventually, moving GitHub Copilot toward token-based billing.

First-order effects

  • At the reported $10 price, GitHub Copilot’s heavier users create losses well above subscription revenue, leaving Microsoft to absorb inference costs or change pricing and usage terms.
  • Developers receive a subsidized AI coding tool in the near term, but their high-volume use is directly at odds with a simple all-you-can-use subscription.

Second-order effects

  • Microsoft has a stronger incentive to steer Copilot toward higher-value enterprise bundles, usage caps, or metered tiers rather than rely solely on the original flat-rate offer.
  • Rival coding-assistant vendors must compete not only on model quality but on whether their pricing covers variable compute costs; aggressive flat rates become harder to sustain.

Third-order effects

  • AI software is likely to be priced increasingly around consumption or measured value as inference becomes a recurring cost of delivery rather than a negligible hosting expense.
  • The durable dividing line in AI applications will be cost per useful task: products that cannot align usage intensity with revenue may need bundling, limits, or cross-subsidy to scale.

The trend: Generative-AI products are moving from adoption-focused flat subscriptions toward pricing models that expose or recover per-use compute costs.

Discussion

  • @trevorcroker Trevor Croker on threads
    Just like the early days of streaming services, we are in the golden age of subsidized AI services.  Eventually, companies will have to see a return on their investments, but the battle for market share is going to bleed profits from some deep pockets for awhile. https://www.wsj.…
  • @jaspar@mastodon.online J Emory Parker on mastodon
    Important warning for anyone experimenting with integrating AI into workflows: The companies selling AI are doing the classic tech thing where they're selling it at a massive loss at the start.  But these models are crazy expensive to run.  Eventually that cost is going to be pas…
  • @cityofthetown Tom Dotan on x
    Microsoft was racking up an OpenAI bill so large on its hit AI product Github Copilot that it was losing more than $20/month per person earlier this year But high costs are a problem for all tech companies trying to turn AI hype into profitable business https://www.wsj.com/...
  • @raamana_ Pradeep Raamana on x
    wow - GitHub Copilot is apparently losing over $20 per month per user (~$80 for some) good example: Using #ChatGPT4 to summarize an email is like getting a Lamborghini to deliver a pizza 🤣🤣 “#AI often doesn't have the economies of scale of standard software because it can...